The study was divided into two phases, as illustrated in Fig. 1. In the first stage, we conducted multivariable logistic regression analyses using data from the NHANES database to examine the connection between liver disorders and periodontitis. In the second stage, we conducted four MR analyses using data from the Gene-Lifestyle Interactions in Dental Endpoints (GLIDE) consortium and FinnGen database to look at the mutual relationship that exists between liver disorders and periodontitis.
NHANES
Data sources and study population
NHANES is a collection of a sequence of cross-sectional surveys aimed at the non-institutionalized population in the US. It selects a nationally representative sample using a multi-stage probability sampling technique and assesses the nutritional and health condition of the participants. The survey includes interviews conducted at participants' households, physical evaluations, and laboratory examinations. NHANES is administered by the National Center for Health Statistics, a division of the Centers for Disease Control and Prevention (CDC). The National Center for Health Statistics Ethics Review Board has given the study ethical approval, and all participants have given written informed permission (Fig. 2).
Figure 2 depicts the process of participant selection for our NHANES study. Initially, there were 30,468 individuals enrolled. Nevertheless, 11,756 participants had to be excluded due to the unavailability of periodontal data. As a result, the study ultimately included a total of 4,425 participants.
Assessment of periodontitis and liver diseases
The oral health statistics from NHANES (2009 to 2014) show that qualified examiners evaluated the clinical attachment loss (CAL) and periodontal probing depth (PD) at six specific locations for every tooth (wisdom teeth excluded), for a total of 28 teeth. The 2018 global classification criteria were used to determine the diagnosis of periodontitis: when interdental CAL is evident in ≥ 2 non-adjacent teeth or when buccal or oral CAL ≥ 3 mm is observed in ≥ 2 teeth with a pocket depth exceeding 3 mm, it is classified as periodontitis (27).
NAFLD was diagnosed by Fatty Liver Index (FLI) ≥ 60 or ultrasonographic FLI (USFLI) ≥ 30. FLI is a numerical number used in medicine to assess a person's risk of having hepatic steatosis or fatty liver disease. Instead of using FLI as a diagnostic sign in our investigation, we opted to employ USFLI. Here is how FLI and USFLI were determined (28):
("Mexican American" and "non-Hispanic Black" have a value of 0 if the participant is not of that ethnicity, and 1 if they are.)
Participants who were at risk of atrial fibrillation (AF) were identified by elevated scores for the non-alcoholic fatty liver disease fibrosis (NFS), fibrosis-4 index (FIB-4), or the aspartate aminotransferase (AST)/platelet ratio index (APRI). The NFS is a scoring system that's used to determine how likely or severe it is that NAFLD may cause fibrosis (the production of scar tissue) in the liver. Without requiring a liver biopsy, medical personnel can assess the degree of fibrosis using the easily accessible and non-invasive FIB-4 instrument. To assess the level of liver fibrosis (scarring) in individuals suffering from liver disease, especially those with chronic hepatitis C, doctors employ the non-invasive APRI scoring system. Here is how these indicators were computed:
(AST: The upper limit of normal (ULN) for the 1999–2000 cycle was 40 U/L; thereafter, the ULN was 33 U/L) (29)
Those with an APRI exceeding 1, FIB-4 greater than 2.67, or NFS surpassing 0.676 were classified as individuals at high risk of AF, as per the criteria outlined in reference.
Assessment of covariates
The thorough evaluation of covariates plays a pivotal role in research, as it serves to control potential confounding factors, thereby ensuring a more precise and dependable assessment of the primary associations. In our study, we have carefully chosen a set of covariates to enhance our comprehension of the link between periodontal disease and liver conditions. We included several key covariates such as age, gender, ethnicity, BMI, drinking, recent tobacco use, history of diabetes, education, marriage, family monthly poverty level index, energy intake and leisure time physical activity (LTPA), all of which were gathered through standardized questionnaires. Additionally, the weight and height of each participant were measured during physical examinations, with BMI calculated as weight (kg)/ height (m2).
Statistical analyses
We combined data from the years 2009 to 2014 and created 6-year sampling weights in accordance with NHANES' sampling methodology, which were incorporated into all of our analyses. To assess variances in continuous and categorical variables across various analysis groups, we used the t-test for continuous data and the Rao Scott chi-square test for categorical variables. The impact of liver disorders on the risk of periodontitis was investigated using a multivariable logistic regression analysis. We computed odds ratios and the associated 95% confidence intervals. In the multivariable analysis, we made the following adjustments: In Model 1, no covariate adjustments were performed; in Model 2, age, gender, and race adjustments were applied; in Model 3, BMI, education level, household income poverty ratio, smoking status, physical activity, and history of diabetes were adjusted. All analyses were performed using SAS 9.4. The significance threshold was set at 0.05 and two-sided levels of significance were computed.
MR Analysis
Basic concept of MR analysis
When compared to traditional observational approaches, MR analysis is less susceptible to errors resulting from reverse causation and confounding since genetic differences are assigned randomly during gamete development and are not connected to environmental variables. As a result, we used MR analysis in this work to find single nucleotide polymorphisms (SNPs) connected to liver disorders and periodontitis. Following this, we integrated these identified SNPs to ascertain the connection between periodontitis and liver diseases.
Study design description
The investigation of the relationship between liver disorders and periodontitis used a two-sample MR analysis. We conducted the MR analyses utilizing summary statistics from open-access databases to examine the association between periodontitis as the exposure and liver conditions including NAFLD, fibrosis, cirrhosis and fibrosis/cirrhosis as the outcomes. The study does not require ethical approval because it is based on publically available summary statistics.
Data sources and selection of instrumental variables (IVs)
The summary statistics derived from two European datasets served as the foundation for the two-sample MR analysis: data for periodontitis (17,353 cases; 28,210 controls) from a genome-wide association study (GWAS) of the European studies of GLIDE consortium (30) ; and liver conditions including NAFLD (2,275 cases; 375,002 controls), fibrosis (146 cases; 373,307 controls), cirrhosis (1,142 cases; 373,307 controls) and fibrosis/cirrhosis (1,841 cases; 375002 controls) from FinnGen (31) which was based on over 370,000 Finnland residents with trustable diagnoses in wide genres of diseases.
SNPs were chosen for IV selection at a genome-wide significance criterion (p < 5 × 10 − 6). SNPs that were deemed eligible for clumping were selected using a 10,000 kb window size and linkage disequilibrium as determined by r2 > 0.001. Palindromic SNPs were excluded from exposure and outcome data when harmonizing them. Finally, we vertificated that all IVs did not impact the outcomes through other pathways by applying Phenoscanner for possible related traits. In order to assure the trustworthiness of results and reduce the interference of confounders related to SNPs with P-value < 1 ×10 − 5, the PhenoScanner database was utilized (32).
Statistical analysis
The random-effects inverse variance weighting (IVW) approach is the main statistical methodology used in this study to investigate the possibility of bidirectional causation between liver illness and periodontitis. To further enhance our results, we also used the weighted mode, weighted median, and MR Egger techniques. The IVW method operates on the premise that all fundamental assumptions of MR are satisfied. Nevertheless, given that the inclusion of pleiotropic IVs can introduce bias into IVW estimates, we conducted sensitivity analyses to account for any pleiotropic effects. R (version 4.3.2) was utilized to conduct the MR analysis. Three packages, "TwoSampleMR"(33), "MRPRESSO"(34), and "forestploter," were specifically used for the data processing and visualization. The study was reported using the STROBE-MR (STrengthening the Reporting of OBservational studies in Epidemiology using Mendelian randomization) criteria (35).
Pleiotropy and Sensitivity Analysis
MR-Egger regression was used to determine if horizontal pleiotropy would exist. The average pleiotropic impact of the IVs is reflected in the intercept term of the MR-Egger regression. In addition, we looked for the existence of pleiotropy using the MR Pleiotropy REsidual Sum and Outlier (MR-PRESSO) test. MR-PRESSO serves multiple purposes, including the detection of horizontal pleiotropy, the correction of horizontal pleiotropy by identifying and removing outliers, and the evaluation of significant differences in causal effects both before and after the elimination of outliers. We used MR-Egger regression and the IVW technique to measure heterogeneity, and we calculated the degree of heterogeneity using the Cochran's Q statistic. We also performed a leave-one-out analysis to assess the consistency and robustness of our results.